TL;DR
The U.S. Commerce Department placed Anthropic’s Claude Fable 5 and Mythos 5 under export controls on June 12, forcing Anthropic to disable both models worldwide. The stated security basis remains disputed, but the shutdown has given AI buyers a concrete example of regulatory switch-off risk.
The U.S. Commerce Department placed Anthropic’s Claude Fable 5 and Mythos 5 under export controls on June 12, forcing the company to disable both models worldwide and turning a product launch into a test case for whether commercial frontier AI can be treated as a national-security export risk.
According to the source material, Commerce Secretary Howard Lutnick sent Anthropic CEO Dario Amodei a letter at 5:21 p.m. ET on June 12 barring access to the two models by any foreign national, including foreign-national employees inside Anthropic. Anthropic said there was no practical way to limit access under those terms while keeping the models online, so it disabled Claude Fable 5 and Mythos 5 for every customer by midnight.
The models had been released three days earlier, on June 9. Fable 5 was described as the public, guard-railed commercial model, while Mythos 5 was the more powerful underlying system made available to selected cyber-defense organizations through Project Glasswing.
The reason for the order is contested. Anthropic said the government gave no specific rationale in the letter and called the action a “misunderstanding.” Reports cited in the source material say officials were alarmed by jailbreak findings, possible cyberattack-usable outputs and concern that a China-linked group may have obtained access.
Washington just switched off
a frontier model
On June 12, an export-control order forced Anthropic to disable Claude Fable 5 and Mythos 5 worldwide. The security merits are still contested. The lesson buyers took away is not: frontier AI can be turned off.
■ The government’s case
- A reported jailbreak pulled malicious, agentic outputs (UK AISI)
- Amazon told officials Fable yielded cyberattack-usable info
- Suspicion a China-linked group obtained the model
- Proliferation & reverse-engineering risk to national security
▲ Anthropic & 120+ experts
- Calls it a narrow, non-universal jailbreak — a “misunderstanding”
- Capability is real but not unique (GPT-5.5, Opus, Kimi 2.7)
- Controls remove tools from defenders, not just attackers
- Export rules built for chips & ore don’t fit software
The precedent is the story. Whatever the jailbreak’s true severity, the U.S. showed it can dark a commercial American model worldwide on ~90 minutes’ notice. Adoption was supposed to be the moat — this week it became the exposure, and the likely winner is the open, sovereign, self-hosted stack.
Shutdown Risk Hits AI Buyers
The immediate cost is not only lost access to two Anthropic systems. The wider effect is that AI customers now have a confirmed case in which a U.S. frontier model was removed from service worldwide on short notice because of an export-control order.
That matters for banks, cloud providers, software companies and government contractors building workflows around hosted AI systems. If a model can be disabled globally because regulators view it as an export risk, buyers may treat single-provider dependence as a business continuity issue, not only a technical choice.
The source material cites Deutsche Bank as saying buyers may now view switch-off risk as proven rather than theoretical. It also points to likely pressure toward multi-model procurement, self-hosted systems and open-weight alternatives that cannot be remotely revoked by one vendor under a government order.
Train It. Tame It. Teach It.: Build Your Personal AI Team and Get Every Model to Work Your Way (The AI Practitioner's Edge)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Security Claims Remain Disputed
The government case, as described in the source material, centers on security risk. The U.K. AI Safety Institute’s red-team lead reportedly said his team found a jailbreak within hours and later extended it to multi-step agentic tool use. The Wall Street Journal reported that Amazon alerted officials after researchers obtained information from Fable 5 that Amazon viewed as usable in cyberattacks.
Semafor reported that the order was also tied in part to suspicion that a China-linked group had obtained the model, creating concern about reverse engineering. David Sacks, identified in the source material as a co-chair of the President’s science-and-tech advisory council, said the White House had asked Amodei the prior week to patch or pull Fable 5 and that Amodei declined.
Anthropic and more than 120 cybersecurity executives and engineers pushed back. Their argument, according to the source material, is that the issue was a narrow jailbreak rather than a universal break, and that comparable security capabilities exist in other frontier and open-weight models.
“misunderstanding”
— Anthropic
Key Facts Still Unsettled
It is not clear from the available source material whether the government has a classified assessment beyond the reported jailbreaks, Amazon’s findings and concern about possible China-linked access. It is also not clear whether the models can be patched in a way that satisfies Commerce while allowing Anthropic to restore service.
The scale of customer disruption has not been fully detailed. The source material says the models were disabled for every customer worldwide, but it does not provide confirmed revenue exposure, contract impact, refund obligations or the number of affected customers.
June 22 Talks Loom
Anthropic and the White House are expected to meet on June 22. The main question is whether the government lifts, narrows or maintains the controls, and whether Anthropic can offer mitigations that allow Fable 5 or Mythos 5 to return in some form.
For the broader AI market, the next test is how customers respond. Procurement teams may add regulatory availability clauses, seek backup providers or move sensitive workloads toward self-hosted and sovereign systems while the Anthropic order is reviewed.
Key Questions
What exactly did the export-control order do?
According to the source material, the order barred access to Claude Fable 5 and Mythos 5 by any foreign national, anywhere. Anthropic said it could not comply cleanly while keeping the models live, so it disabled both worldwide.
Why did the U.S. government act against the models?
The reason has not been fully confirmed. Reports cited in the source material point to jailbreak findings, cyberattack-related concerns raised by Amazon and suspicion of China-linked access. Anthropic said the directive did not give a specific rationale.
Does this mean the models were proven unsafe?
No public record in the source material establishes that. Anthropic said the reported issue was narrow and not a universal jailbreak. Government-linked and media accounts describe serious concerns, but the full technical record has not been released.
Why are AI customers alarmed?
The order showed that access to a hosted frontier model can be removed globally on short notice. That may push customers to spread workloads across providers or favor systems they can run under their own control.
When could the situation change?
The next known milestone is the expected June 22 meeting between Anthropic and the White House. Any change would depend on whether officials accept Anthropic’s security response or revise the export restrictions.
Source: Thorsten Meyer AI
